Computer clinical decision support that automates personalized clinical care: a challenging but needed healthcare delivery strategy.

How to deliver best care in various clinical settings remains a vexing problem. All pertinent healthcare-related questions have not, cannot, and will not be addressable with costly time- and resource-consuming controlled clinical trials. At present, evidence-based guidelines can address only a small fraction of the types of care that clinicians deliver. Furthermore, underserved areas rarely can access state-of-the-art evidence-based guidelines in real-time, and often lack the wherewithal to implement advanced guidelines. Care providers in such settings frequently do not have sufficient training to undertake advanced guideline implementation. Nevertheless, in advanced modern healthcare delivery environments, use of eActions (validated clinical decision support systems) could help overcome the cognitive limitations of overburdened clinicians. Widespread use of eActions will require surmounting current healthcare technical and cultural barriers and installing clinical evidence/data curation systems. The authors expect that increased numbers of evidence-based guidelines will result from future comparative effectiveness clinical research carried out during routine healthcare delivery within learning healthcare systems.

Enabling a learning healthcare system with automated computer protocols that produce replicable and personalized clinician actions.

Clinical decision-making is based on knowledge, expertise, and authority, with clinicians approving almost every intervention-the starting point for delivery of "All the right care, but only the right care," an unachieved healthcare quality improvement goal. Unaided clinicians suffer from human cognitive limitations and biases when decisions are based only on their training, expertise, and experience. Electronic health records (EHRs) could improve healthcare with robust decision-support tools that reduce unwarranted variation of clinician decisions and actions. Current EHRs, focused on results review, documentation, and accounting, are awkward, time-consuming, and contribute to clinician stress and burnout. Decision-support tools could reduce clinician burden and enable replicable clinician decisions and actions that personalize patient care. Most current clinical decision-support tools or aids lack detail and neither reduce burden nor enable replicable actions. Clinicians must provide subjective interpretation and missing logic, thus introducing personal biases and mindless, unwarranted, variation from evidence-based practice. Replicability occurs when different clinicians, with the same patient information and context, come to the same decision and action. We propose a feasible subset of therapeutic decision-support tools based on credible clinical outcome evidence: computer protocols leading to replicable clinician actions (eActions). eActions enable different clinicians to make consistent decisions and actions when faced with the same patient input data. eActions embrace good everyday decision-making informed by evidence, experience, EHR data, and individual patient status. eActions can reduce unwarranted variation, increase quality of clinical care and research, reduce EHR noise, and could enable a learning healthcare system.

Percentage of Time in Range 70 to 139 mg/dL Is Associated With Reduced Mortality Among Critically Ill Patients Receiving IV Insulin Infusion.

BACKGROUND: In addition to hyperglycemia, hypoglycemia, and glycemic variability, reduced time in targeted blood glucose range (TIR) is associated with increased risk of death in critically ill patients. This relation between TIR and mortality may be confounded by diabetic status and antecedent glycemic control. METHODS: This study retrospectively analyzed critically ill patients managed with the same IV insulin protocol at multiple centers. The percentage of TIR between 70 and 139 mg/dL was calculated. Patients with diabetic ketoacidosis, patients who had < 10 blood glucose readings, and patients with repeat admissions were excluded. The highest recorded glycosylated hemoglobin value in the preceding 3 months or up to 1 month following admission were used as a surrogate for the patient's preexisting glucose control. Stratified regression analyses were performed for 30-day mortality, with covariates of age, sex, TIR ≥ 80%, Acute Physiology Score, and Charlson Comorbidity Index. RESULTS: A total of 9,028 patients, 53.2% of whom had diabetes, were studied. Median TIR was 84.1% for nondiabetic patients and 64.5% for patients with diabetes. Mortality was lower in those with TIR > 80% compared with those with TIR ≤ 80% (12.4% vs 19.2%; P < .001). TIR > 80% was independently associated with reduced mortality in nondiabetic patients (OR, 0.52; P < .001), patients with diabetes (OR, 0.69; P = .001), and patients with well-controlled disease (OR, 0.50; P < .001) but not in patients with poorly controlled disease (OR, 0.86; P = .40). CONCLUSIONS: TIR was independently associated with mortality in critically ill patients, particularly those with good antecedent glucose control.

Airflow Obstruction Categorization Methods and Mortality.

RATIONALE: Current guidelines recommend using forced expiratory volume in 1 second (FEV1) % predicted to categorize the severity of airflow obstruction. There are limitations to using FEV1 % predicted for this purpose, including bias associated with demographic factors and the inability to correct for "lung size." Other methods for grading the severity of airflow obstruction have been proposed to address these limitations. OBJECTIVES: Our objectives were to categorize airflow obstruction severity using these methods and then determine which method results in a categorization most closely associated with mortality. METHODS: Study subjects were patients aged 40-80 years tested in our pulmonary function test laboratories in the period 2002 to 2013 with airflow obstruction based on an FEV1/forced vital capacity (FVC) less than the lower limit of normal. Categorization of airflow obstruction severity was determined using four methods: FEV1 % predicted; FEV1 % predicted adjusted by FVC % predicted; FEV1/FVC confidence interval approach; and FEV1 z-scores. Receiver operating characteristic curve analysis was used to determine which categorization method best predicts 5-year survival. RESULTS: We identified 2,000 patients with airflow obstruction. Important differences in the categorization of airflow obstruction severity were observed using the different methods. More patients were categorized as having severe obstruction using FEV1 % predicted and FEV1 z-scores compared with FEV1 % predicted adjusted by FVC % predicted and FEV1/FVC confidence interval approach. FEV1 % predicted was the best predictor of 5-year survival among the four methods studied. CONCLUSIONS: In our study, categorizing airflow obstruction severity using FEV1 % predicted best predicted 5-year survival. This validates the current guideline recommendation that FEV1 % predicted be used to categorize the severity of airflow obstruction.

Lower Glucose Target Is Associated With Improved 30-Day Mortality in Cardiac and Cardiothoracic Patients.

BACKGROUND: Practice guidelines recommend against intensive insulin therapy in patients who are critically ill based on trials that had high rates of severe hypoglycemia. Intermountain Healthcare uses a computerized IV insulin protocol that allows choice of blood glucose (BG) targets (80-110 vs 90-140 mg/dL) and has low rates of severe hypoglycemia. We sought to study the effects of BG target on mortality in adult patients in cardiac ICUs that have very low rates of severe hypoglycemia. METHODS: Critically ill patients receiving IV insulin were treated with either of two BG targets (80-110 vs 90-140 mg/dL). We created a propensity score for BG target using factors thought to have influenced clinicians' choice, and then we performed a propensity score-adjusted regression analysis for 30-day mortality. RESULTS: There were 1,809 patients who met inclusion criteria. Baseline patient characteristics were similar. Median glucose was lower in the 80-110 mg/dL group (104 vs 122 mg/dL, P < .001). Severe hypoglycemia occurred at very low rates in both groups (1.16% vs 0.35%, P = .051). Unadjusted 30-day mortality was lower in the 80-110 mg/dL group (4.3% vs 9.2%, P < .001). This remained after propensity score-adjusted regression (OR, 0.65; 95% CI, 0.43-0.98; P = .04). CONCLUSIONS: Tight glucose control can be achieved with low rates of severe hypoglycemia and is associated with decreased 30-day mortality in a cohort of largely patients in cardiac ICUs. Although such findings should not be used to guide clinical practice at present, the use of tight glucose control should be reexamined using a protocol that has low rates of severe hypoglycemia.

Human Cognitive Limitations. Broad, Consistent, Clinical Application of Physiological Principles Will Require Decision Support.

Our education system seems to fail to enable clinicians to broadly understand core physiological principles. The emphasis on reductionist science, including "omics" branches of research, has likely contributed to this decrease in understanding. Consequently, clinicians cannot be expected to consistently make clinical decisions linked to best physiological evidence. This is a large-scale problem with multiple determinants, within an even larger clinical decision problem: the failure of clinicians to consistently link their decisions to best evidence. Clinicians, like all human decision-makers, suffer from significant cognitive limitations. Detailed context-sensitive computer protocols can generate personalized medicine instructions that are well matched to individual patient needs over time and can partially resolve this problem.

A Pediatric Intensive Care Unit Bedside Computer Clinical Decision Support Protocol for Hyperglycemia Is Feasible, Safe and Offers Advantages.

BACKGROUND: Computer clinical decision support (CDS) systems are uncommon in the pediatric intensive care unit (PICU), despite evidence suggesting they improve outcomes in adult ICUs. We reasoned that a bedside CDS protocol for intravenous insulin titration, eProtocol-insulin, would be feasible and safe in critically ill children. METHODS: We retrospectively reviewed data from non-diabetic children admitted to the PICU with blood glucose (BG) ≥140 mg/dL who were managed with intravenous insulin by either unaided clinician titration or eProtocol-insulin. Primary outcomes were BG measurements in target range (80-110 mg/dL) and severe hypoglycemia (BG ≤40 mg/dL); secondary outcomes were 60-day mortality and PICU length of stay. We assessed bedside nurse satisfaction with the eProtocol-insulin protocol by using a 5-point Likert scale and measured clinician compliance with eProtocol-insulin recommendations. RESULTS: Over 5 years, 69 children were titrated with eProtocol-insulin versus 104 by unaided clinicians. eProtocol-insulin achieved target range more frequently than clinician titration (41% vs. 32%, P < 0.001). Severe hypoglycemia was uncommon in both groups (4.3% of patients in eProtocol-insulin, 8.7% in clinician titration, P = 0.37). There were no differences in mean time to BG target or median BG between the groups. Mortality was 23% in both groups. Clinician compliance with eProtocol-insulin recommendations was 89%. Nurses believed that eProtocol-insulin was easy to understand and safer than clinician titration. CONCLUSIONS: eProtocol-insulin is safe for titration of intravenous insulin in critically ill children. Clinical research protocols and quality improvement initiatives aimed at optimizing BG control should utilize detailed computer protocols that enable replicable clinician decisions.

Evaluating How Post-Bronchodilator Vital Capacities Affect the Diagnosis of Obstruction in Pulmonary Function Tests.

BACKGROUND: Although the ratio of FEV1 to the vital capacity (VC) is universally accepted as the cornerstone of pulmonary function test (PFT) interpretation, FVC remains in common use. We sought to determine what the differences in PFT interpretation were when the largest measured vital capacity (VCmax) was used instead of the FVC. METHODS: We included 12,238 consecutive PFTs obtained for routine clinical care. We interpreted all PFTs first using FVC in the interpretation algorithm and then again using the VCmax, obtained either before or after administration of inhaled bronchodilator. RESULTS: Six percent of PFTs had an interpretive change when VCmax was used instead of FVC. The most common changes were: new diagnosis of obstruction and exclusion of restriction (previously suggested by low FVC without total lung capacity measured by body plethysmography). A nonspecific pattern occurred in 3% of all PFT interpretations with FVC. One fifth of these 3% produced a new diagnosis of obstruction with VCmax. The largest factors predicting a change in PFT interpretation with VCmax were a positive bronchodilator response and the administration of a bronchodilator. Larger FVCs decreased the odds of PFT interpretation change. Surprisingly, the increased numbers of PFT tests did not increase odds of PFT interpretation change. CONCLUSIONS: Six percent of PFTs have a different interpretation when VCmax is used instead of FVC. Evaluating borderline or ambiguous PFTs using the VCmax may be informative in diagnosing obstruction and excluding restriction.

Comparison of NHANES III and ERS/GLI 12 for airway obstruction classification and severity.

The diagnosis and severity categorisation of obstructive lung disease is determined using reference values. The American Thoracic Society/European Respiratory Society in 2005 recommended the National Health and Nutrition Examination Survey (NHANES) III spirometry prediction equations for patients in USA aged 8-80 years. The Global Lung Initiative 2012 (GLI 12) provided spirometry prediction equations for patients aged 3-95 years. Comparison of the NHANES III and GLI 12 prediction equations for diagnosing and categorising airway obstruction in patients in USA has not been made.We aimed to quantify the differences between NHANES III and GLI 12 predicted values in Caucasians aged 18-95 years, using both mathematical simulation and clinical data. We compared predicted forced expiratory volume in 1 s (FEV1) and lower limit of normal (LLN) FEV1/forced vital capacity (FVC) % for NHANES III and GLI 12 prediction equations by applying both a simulation model and clinical spirometry data to quantify differences in the diagnosis and categorisation of airway obstruction.Mathematical simulation revealed significant similarities and differences between prediction equations for both LLN FEV1/FVC % and predicted FEV1 There are significant differences when using GLI 12 and NHANES III to diagnose airway obstruction and severity in Caucasian patients aged 18-95 years.Similarities and differences exist between NHANES III and GLI 12 for some age and height combinations. The differences in LLN FEV1/FVC % and predicted FEV1 are most prominent in older taller/shorter individuals. The magnitude of the differences can be large and may result in differences in clinical management.

An in silico method to identify computer-based protocols worthy of clinical study: An insulin infusion protocol use case.

OBJECTIVE: Develop an efficient non-clinical method for identifying promising computer-based protocols for clinical study. An in silico comparison can provide information that informs the decision to proceed to a clinical trial. The authors compared two existing computer-based insulin infusion protocols: eProtocol-insulin from Utah, USA, and Glucosafe from Denmark. MATERIALS AND METHODS: The authors used eProtocol-insulin to manage intensive care unit (ICU) hyperglycemia with intravenous (IV) insulin from 2004 to 2010. Recommendations accepted by the bedside clinicians directly link the subsequent blood glucose values to eProtocol-insulin recommendations and provide a unique clinical database. The authors retrospectively compared in silico 18,984 eProtocol-insulin continuous IV insulin infusion rate recommendations from 408 ICU patients with those of Glucosafe, the candidate computer-based protocol. The subsequent blood glucose measurement value (low, on target, high) was used to identify if the insulin recommendation was too high, on target, or too low. RESULTS: Glucosafe consistently provided more favorable continuous IV insulin infusion rate recommendations than eProtocol-insulin for on target (64% of comparisons), low (80% of comparisons), or high (70% of comparisons) blood glucose. Aggregated eProtocol-insulin and Glucosafe continuous IV insulin infusion rates were clinically similar though statistically significantly different (Wilcoxon signed rank test P = .01). In contrast, when stratified by low, on target, or high subsequent blood glucose measurement, insulin infusion rates from eProtocol-insulin and Glucosafe were statistically significantly different (Wilcoxon signed rank test, P < .001), and clinically different. DISCUSSION: This in silico comparison appears to be an efficient nonclinical method for identifying promising computer-based protocols. CONCLUSION: Preclinical in silico comparison analytical framework allows rapid and inexpensive identification of computer-based protocol care strategies that justify expensive and burdensome clinical trials.

Clinical study replicability and the pursuit of excellence.

Comparisons of processes of care are common in critical care research. Often, these processes are neither explicit nor replicable and this can result in seemingly irreconcilable results. Here, we briefly review the article by Taniguchi and colleagues, who studied liberation from mechanical ventilation by using either a computerized weaning protocol or one driven by respiratory therapists. We discuss the implications of explicit protocols increasing replicability in clinical research.

Standardizing Predicted Body Weight Equations for Mechanical Ventilation Tidal Volume Settings.

BACKGROUND: Recent recommendations for lung protective mechanical ventilation include a tidal volume target of 6 mL/kg predicted body weight (PBW). Different PBW equations might introduce important differences in tidal volumes delivered to research subjects and patients. METHODS: PBW equations use height, age, and sex as input variables. We compared National Institutes of Health (NIH) ARDS Network (ARDSNet), actuarial table (ACTUARIAL), and Stewart (STEWART) PBW equations used in clinical trials, across physiologic ranges for age and height. We used three-dimensional and two-dimensional surface analysis to compare these PBW equations. We then used age and height from actual clinical trial subjects to quantify PBW equation differences. RESULTS: Significant potential differences existed between these PBW predictions. The ACTUARIAL and ARDSNet surfaces for women were the only surfaces that intersected and produced both positive and negative differences. Mathematical differences between PBW equations at limits of height and age exceeded 30% in women and 24% in men for ACTUARIAL vs ARDSNet and about 25% for women and 15% for men for STEWART vs ARDSNet. The largest mathematical differences were present in older, shorter subjects, especially women. Actual differences for clinical trial subjects were as high as 15% for men and 24% for women. CONCLUSIONS: Significant differences between PBW equations for both men and women could be important sources of interstudy variation. Studies should adopt a standard PBW equation. We recommend using the NIH National Heart, Lung, and Blood Institute ARDS Network PBW equation because it is associated with the clinical trial that identified 6 mL/kg PBW as an appropriate target.